Software Everywhere — David Beers on Mobile Computing

Will ALP run X Windows? That's a question that a lot of Linux developers will want to know because it will have implications for how easy it will be to port existing Linux applications to the new version of "Palm OS". For example, it would probably be a lot easier to port a Skype clone like Gizmo onto an ALP device if the platform had an X Server handling the various windowing tasks. Handheld.org's GPE and Nokia's Maemo platforms, based on the same GTK+ toolkit as ALP, run X Windows under the hood. And according to PalmSource developer Jesse Donaldson the demo PalmSource gave in Barcelona this week was with a version of their MAX framework that uses X.

But X Windows isn't the only way to use GTK+ as your GUI toolkit. GTK+ can also work directly with the framebuffer using DirectFB or possibly be ported directly to the low-level graphics APIs to remove the intermediate interface that X Windows provides to graphics. Jesse says that the jury is still out at PalmSource as to whether X Windows is worth the performance hit.

One school of thought would say that by the time ALP reaches real devices (if we're lucky before the end of 2007) Moore's Law will have accommodated enough of the weight of X Windows that the slow performance observed on devices like the Nokia 770 Internet Tablet won't be such a concern. (In fairness, there's more than just X Windows going on with the 770's Maemo platform: Nokia has overlaid their own UI framework that they call the Hildon UI, which itself could account for some of the 770's less-than-snappy performance.)

Perhaps.

But I'm of the mind that what ALP needs is not so much an easy migration path for desktop Linux applications to the handset as it is a truly great user experience that we haven't seen on mobile Linux yet. Instantaneous response to user input has always been a hallmark that differentiated the Palm OS from the likes of Windows CE and it accounts for some considerable part of Palm's popularity, in my opinion.

Software performance also trickles down to choices of processors, which in turn affect cost, battery life, the physical size of the device--all stuff that makes a big difference in the success of a handset. I think the Treo is a remarkable piece of engineering and a triumph of good design, but let's face it: it would have a vastly greater market if it could accomplish what it does with a smaller battery and have a form factor more like the Motorola RAZR. The gains from Moore's Law should be spent on reducing the physical size of mobile computers and smartphones, not supporting bloated software.

Now, I'm oversimplifying the issue by describing X Windows as if it was nothing but bloat. I'm sure that if PalmSource has to build its MAX framework on the direct-to-framebuffer version of GTK+ it will mean they have to write more of their own code to handle things that X Windows would have been doing for them. And there are probably plenty of considerations here that I can't judge out of ignorance. But in weighing the trade-offs I'd like to urge PalmSource to put a high priority on performance and to take a leadership roll in paring Linux down to the essentials needed for mobile devices.